DETECTION OF OUTBOUND SENDING OF SPAM

• US 20100077040A1
• Filed: 09/17/2009
• Published: 03/25/2010
• Est. Priority Date: 09/19/2008
• Status: Active Grant

• Current Assignee: Yahoo Assets LLC
• Original Assignee: Yahoo! Inc. (Apollo Global Management, Inc.)
• Inventors: Kundu, Anirban, Zheng, Lei, Ramarao, Vishwanath Tumkur, Risher, Mark E., Xi, Xiaopeng, Jamal, Naveen, Hariharan, Bhasker
• Granted Patent: US 8,868,663 B2
• Time in Patent Office: Days
• US Class Current: 709/206
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/01 Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
• CPC Class Codes: H04L 51/212 using filtering or selective blocking
• CPC Class Codes: H04L 63/1425 Traffic logging, e.g. anomaly detection

Abstract

The invention provides for at least three processes for detecting the probability of abusive use of a message account for sending large amounts of unsolicited messages, such as spam, to other message accounts. For example, information provided at registration for a new message account can be processed to determine the likelihood of abusive use of that message account. Also, inbound messages can be processed to determine if the message account that sent the inbound message is abusing the use of that message account. Additionally, outbound messages can be processed to determine if the message account that is attempting to send an outbound message is abusing the use of that message account. Each of these three processes can operate separately or in any combination with each other to further improve the probability that abusive use of a message account will be detected promptly and accurately.

Claims

1. A network device, comprising:

a transceiver to send and receive data over a network; and

a processor that receives data from and sends data to the transceiver, and performs actions, including;

intercepting a message that is sent outbound from a message account to a message destination;

parsing the outbound message to identify a plurality of message characteristics;

analyzing the plurality of message characteristics to determine a plurality of probability values;

assigning a probability score to the outbound message based in part on the determined plurality of probability values and previously determined data and probability scores for a plurality of other message accounts;

selectively classifying the outbound message as abusive based on the assigned probability score being at or above a determined threshold value; and

if the outbound message is classified as abusive, inhibiting sending of the outbound message over the network;

otherwise, enabling the outbound message to be sent towards the message destination.

2. The network device of claim 1, wherein analyzing the plurality of message characteristics further comprises:

determining a message username from the plurality of message characteristics or the message account;

analyzing the message username to determine at least one or more of a length of the username, an amount of transitions in the username, an amount of numbers in the username, an amount of letters in the username, or an amount of symbols in the username; and

based on the analysis of the message username, employing at least one rule to assign a probability value to the outbound message.

3. The network device of claim 1, wherein analyzing the plurality of message characteristics further comprises:

determining a network domain corresponding to the outbound message as at least one message characteristic in the plurality of message characteristics; and

determining a probability value based on the determined network domain, including at least based on whether the determined network domain is an unknown network domain or based on a relationship between the determined network domain and a network domain associated with the message destination.

4. The network device of claim 1, wherein analyzing the plurality of message characteristics further comprises:

analyzing a velocity of similar outbound messages both from the message account associated with the outbound message and from other message accounts previously classified as abusive message accounts; and

determining a probability value based on the analysis of the velocity.

5. The network device of claim 1, wherein analyzing the plurality of message characteristics further comprises:

analyzing a content of similar outbound messages both from the same message account associated with the outbound message and from other message accounts previously classified as abusive message accounts; and

determining a probability value based on the analysis of the content.

6. The network device of claim 1, wherein the processor performs actions, further including:

retrieving a plurality of characteristics and probability scores associated with a plurality of other message account registrations;

determining a degree of similarity of a message account registration for the message account associated with the outbound message and a plurality of other message account registrations that are identified as abusive message accounts; and

based on the degree of similarity, adjusting the probability score to the outbound message.

7. A processor readable storage medium that includes data and instructions, wherein the execution of the instructions on a computing device by enabling actions, comprising:

intercepting a message that is sent outbound from a message account for delivery to a message destination;

parsing the outbound message to identify a plurality of message characteristics;

analyzing the plurality of message characteristics to determine a plurality of probability values;

assigning a probability score to the outbound message based in part on the determined probability values and previously determined data and probability scores for a plurality of other message accounts;

adjusting the probability score for the outbound message based in part on contributing factors indicating whether the message account related to the outbound message was previously used for abusive or fraudulent purposes;

classifying the outbound message as abusive based on the assigned probability score being at or above a determined threshold value; and

if the outbound message is classified as abusive, inhibiting delivery of the outbound message;

otherwise, enabling the outbound message to be delivered to the message destination.

8. The processor readable storage medium of claim 7, wherein adjusting the probability score further comprises:

retrieving message account registration information associated with the outbound message;

retrieving a plurality of characteristics and probability scores associated with a plurality of other message account registrations;

clustering at least a subset of the plurality of characteristics and probability scores associated with the plurality of other message account registrations based in part on a determined similarity in characteristics between the plurality of other message account registrations and the message account registration information for the outbound message;

determining as adjustment for the probability score of the outbound message based on the clustering of at least the subset of plurality of characteristics and probability scores; and

applying the adjustment to the assigned probability score for the outbound message.

9. The processor readable storage medium of claim 7, wherein adjusting the probability score further comprises employing at least one of a machine learning, artificial intelligence or logical decision tree analysis to adjust the probability score of the outbound message.

10. The processor readable storage medium of claim 7, wherein adjusting the probability score further comprises:

wherein the determined contributing factors include at least one of a feature of an application used for registration of the message account associated with the outbound message; and

adjusting the probability score for the outbound message based on a probability value associated with the application used.

11. The processor readable storage medium of claim 7, wherein the contributing factors further comprises at least one of a similarity of contact information for the message account associated with the outbound message and other message accounts.

12. The processor readable storage medium of claim 7, wherein the contributing factors further comprise at least one of an amount of previous outbound messages from the message account associated with the outbound message, or an amount of recipients of the outbound message sent by the associated message account.

13. The processor readable storage medium of claim 7, wherein the instructions enable actions, further including:

employing the contributing factors, the adjusted probability score, or the plurality of message characteristics to update a probability score for at least one other message account.

14. A system for enabling a communications over a network, comprising:

a data storage device having stored thereon, a plurality of abusive message account data from a plurality of message account registrations; and

one or more processors configured to employ data from the data storage device to perform actions, including;

intercepting an outbound message sent from a message account towards a message destination;

parsing the outbound message to identify a plurality of message characteristics;

analyzing the plurality of message characteristics to determine a plurality of probability values;

assigning a probability score to the outbound message based in part on the determined probability values and previously determined data and probability scores for the plurality of other message account registrations;

classifying the outbound message as abusive based on the assigned probability score being at or above a determined threshold value; and

if the outbound message is classified as abusive, inhibiting delivery of the outbound message;

otherwise, enabling the outbound message to be delivered to the message destination.

15. The system of claim 14, wherein the one or more processors perform actions, further including:

determining a plurality of contributing factors from the plurality of abusive message account data from the plurality of message account registrations;

determining a plurality of other contributing factors based on message account registration information associated with the outbound message;

comparing the determined plurality of contributing factors with the plurality of other contributing factors; and

adjusting the probability score based the comparison.

16. The system of claim 14, wherein parsing the outbound message further comprises:

determining a message username associated with the message account;

parsing a plurality of characters in the message username to determine various statistics about the message username including at least one of a length of the message username based on the plurality of characters, a number of transitions between the plurality of characters, a number of digits, a number of letters, or a number of symbols; and

employing one or more rules to assign the probability score based on the statistics about the message username.

17. The system of claim 16, wherein employing the one or more rules, further comprises:

clustering a plurality of characteristics about a username associated with the outbound message and probability scores associated with the plurality of other message account registrations based in part on a determined similarity in characteristics between the plurality of other message account registrations and message account registration information for the outbound message;

determining as adjustment for the probability score of the outbound message based on the clustering of at least the subset of plurality of characteristics and probability scores; and

applying the adjustment to the assigned probability score for the outbound message.

18. The system of claim 14, wherein if the outbound message is classified as abusive, further classifying the message account associated with the outbound message as abusive.

19. The system of claim 14, wherein analyzing the plurality of message characteristics further comprises:

determining a network domain corresponding to the outbound message as at least one message characteristic in the plurality of message characteristics; and

determining a probability value based on the determined network domain, including at least based on whether the determined network domain is an unknown network domain and a relationship between the determined network domain and a network domain associated with the message destination.

20. The system of claim 14, wherein the classification of the outbound message is revised based on feedback from a classification of another message account.